Core Scientific plans $1.8B AI run-rate, Bitcoin mining exit by Q4

Core Scientific expects its colocation business to reach an average annualized GAAP revenue run-rate of approximately $1.8 billion, according to its Q3 2026 earnings report issued November 5, 2026, as it prepares to exit Bitcoin mining by Q4 2026.

The company is shifting from a Bitcoin mining operator to a long-term host for AI infrastructure. Core Scientific described the figure as an “average annualized GAAP revenue run-rate of approximately $1.8 billion,” underscoring that the estimate depends on long-term leases and fully operational facilities. At full operation, it equals roughly $150 million per month, not cash already in hand.

How much of the AI business is already billing?

Core Scientific reported $215.4 million in colocation revenue for the quarter, equal to 92% of $232.9 million in total revenue. The company also posted nearly 60% sequential revenue growth. Billing capacity stood at 485 MW as of September 30, 2026.

A simple annualization of the quarter’s colocation revenue produces $861.6 million, less than half of the $1.8 billion target. The 485 MW billing base also remains less than half of roughly 1.1 GW under contract. AI hosting already supplies most revenue, but the forecast requires Core Scientific to commission facilities and convert reserved capacity into billable service.

What does the 1.1 GW pipeline contain?

Core Scientific has approximately 1.1 GW of contracted power capacity, distributed across two major relationships:

• 590 MW under a long-term deal with CoreWeave.
• 530 MW under a 15-year agreement with an AMD-backed neocloud counterparty.

Core Scientific projects initial deliveries for the AMD-related agreements in the first half of 2027. The two relationships account for the bulk of the pipeline, concentrating execution risk in facility delivery schedules and the performance of two counterparties.

The 590 MW CoreWeave commitment and 530 MW AMD-backed commitment each exceed the 485 MW current billing base, making delivery timing central to the forecast. The gap between contracted and operating capacity also means that the $1.8 billion run-rate depends on data-center facilities entering service rather than on demand for additional contracts alone.

What does the full footprint imply for earnings and credit?

Management expects “approximately $1.6 billion in net operating income” from the full colocation footprint, implying margins of approximately 90%. The company says more than half of expected cash flows should be backed by investment-grade credit.

Core Scientific projects potential colocation revenue of more than $24 billion over the life of its contracts. That figure represents future contractual revenue, not current funds or guaranteed near-term cash. The distinction matters because the $1.8 billion run-rate assumes fully operational facilities, while the first AMD-related deliveries do not begin until 2027.

What does the mining exit change for Core Scientific?

Starting in 2027, Core Scientific intends to operate exclusively as a high-density colocation provider for AI workloads. Its operating focus will shift from mining fleet output to delivering AI capacity against long-term leases and service commitments. The company plans to complete its Bitcoin mining exit by Q4 2026.

For customers, contracted AI capacity provides multi-year access to powered data-center capacity, making delivery timing central to infrastructure planning. For capital providers, contracted revenue and credit quality now carry more analytical weight than Bitcoin mining metrics. Core Scientific itself must raise billable capacity from 485 MW toward roughly 1.1 GW while meeting the margin and credit assumptions behind the forecast.

The concentration in CoreWeave and the AMD-backed counterparty also makes facility delivery and counterparty performance the central execution variables. The company has already crossed the revenue-mix threshold, with colocation accounting for 92% of Q3 sales, but the remaining capacity gap leaves substantial operational risk before the model reaches steady state.

What must happen before the forecast becomes billable?

Three milestones provide the clearest scorecard:

• Complete the Bitcoin mining exit by Q4 2026.
• Increase billing capacity from 485 MW toward roughly 1.1 GW.
• Deliver the first AMD-related capacity in the first half of 2027.

The next hard test arrives by Q4 2026, when the mining exit is due; first AMD-related deliveries in the first half of 2027 will then show whether contracted power can become billable AI colocation on the schedule required for the $1.8 billion target.